Have IDEA's mul() return the result instead of altering a reference value.

Split up some lines for readability. Benchmarks somewhat slower than the
previous version (34.3 vs 32.0 on my Core2, gcc 4.1.2), will need to
investigate.

1 files changed, 25 insertions, 13 deletions

diff --git a/src/idea.cpp b/src/idea.cpp
index 267a14504..49c064073 100644
--- a/src/idea.cpp
+++ b/src/idea.cpp
@@ -13,17 +13,17 @@ namespace {
 /*************************************************
 * Multiplication modulo 65537                    *
 *************************************************/
-inline void mul(u16bit& a, u16bit b)
+inline u16bit mul(u16bit a, u16bit b)
    {
    if(a && b)
       {
       u32bit temp = static_cast<u32bit>(a) * b;
       a = static_cast<u16bit>(temp >> 16);
       b = static_cast<u16bit>(temp & 0xFFFF);
-      a = static_cast<u16bit>(b - a + ((b < a) ? 1 : 0));
+      return static_cast<u16bit>(b - a + ((b < a) ? 1 : 0));
       }
    else
-      a = static_cast<u16bit>(1 - a - b);
+      return static_cast<u16bit>(1 - a - b);
    }
 
 }
@@ -40,24 +40,30 @@ void IDEA::enc(const byte in[], byte out[]) const
 
    for(u32bit j = 0; j != 8; ++j)
       {
-      mul(X1, EK[6*j+0]);
+      X1 = mul(X1, EK[6*j+0]);
       X2 += EK[6*j+1];
       X3 += EK[6*j+2];
-      mul(X4, EK[6*j+3]);
+      X4 = mul(X4, EK[6*j+3]);
+
       u16bit T0 = X3;
       X3 ^= X1;
-      mul(X3, EK[6*j+4]);
+      X3 = mul(X3, EK[6*j+4]);
+
       u16bit T1 = X2;
       X2 = static_cast<u16bit>((X2 ^ X4) + X3);
-      mul(X2, EK[6*j+5]);
+      X2 = mul(X2, EK[6*j+5]);
       X3 += X2;
+
       X1 ^= X2;
       X4 ^= X3;
       X2 ^= T0;
       X3 ^= T1;
       }
 
-   mul(X1, EK[48]); X2 += EK[50]; X3 += EK[49]; mul(X4, EK[51]);
+   X1  = mul(X1, EK[48]);
+   X2 += EK[50];
+   X3 += EK[49];
+   X4  = mul(X4, EK[51]);
 
    store_be(out, X1, X3, X2, X4);
    }
@@ -74,24 +80,30 @@ void IDEA::dec(const byte in[], byte out[]) const
 
    for(u32bit j = 0; j != 8; ++j)
       {
-      mul(X1, DK[6*j+0]);
+      X1 = mul(X1, DK[6*j+0]);
       X2 += DK[6*j+1];
       X3 += DK[6*j+2];
-      mul(X4, DK[6*j+3]);
+      X4 = mul(X4, DK[6*j+3]);
+
       u16bit T0 = X3;
       X3 ^= X1;
-      mul(X3, DK[6*j+4]);
+      X3 = mul(X3, DK[6*j+4]);
+
       u16bit T1 = X2;
       X2 = static_cast<u16bit>((X2 ^ X4) + X3);
-      mul(X2, DK[6*j+5]);
+      X2 = mul(X2, DK[6*j+5]);
       X3 += X2;
+
       X1 ^= X2;
       X4 ^= X3;
       X2 ^= T0;
       X3 ^= T1;
       }
 
-   mul(X1, DK[48]); X2 += DK[50]; X3 += DK[49]; mul(X4, DK[51]);
+   X1 = mul(X1, DK[48]);
+   X2 += DK[50];
+   X3 += DK[49];
+   X4 = mul(X4, DK[51]);
 
    store_be(out, X1, X3, X2, X4);
    }